Olink_R_analysis.qmd

---

Introduction

1. Platform Overview: Olink Explore HT

Olink High-Throughput (HT) proteomic platform that combines the specificity of Proximity Extension Assay (PEA) with Next-Generation Sequencing (NGS) readout.

Key Features

• Multiplexing: Simultaneously measures 5,420 protein biomarkers organized into 8 blocks ExploreHT_Validation.pdf.
• Dynamic Range: Achieves a 10-log dynamic range by utilizing different sample dilutions within the blocks.
• Sample Efficiency: Requires only 2 µL of sample for the entire panel.

---

2. Core Terminology & Data Structure

• Projects: A set of plates that run at the same time and have been normalized together. If two projects are not randomized or are run at different times then additional normalization is required. A project encapsulates all metadata and run results for a specific study. Each project consists of one or more plates. NPX_Manual.pdf.
• Plates: The physical processing units, typically utilizing a standard 96-well format.
• Assays: Individual antibody-based tests designed for specific protein targets. The Olink Explore HT system, for example, features 5,420 protein biomarkers (assays) organized into 8 blocks. Each block includes internal controls: one incubation control, one extension control, and one amplification control. This results in a total of 5,444 assays per sample.
  
• Normalization: The conversion of raw NGS counts into NPX (Normalized Protein Expression), a relative \(log_2\) scale.
  • Plate Control (PC) Normalization: PC normalization is the standard “baseline” normalization. It uses internal controls (Plate Controls) included in every run to account for technical variation between different plates. \[NPX_{i,j} = ExtNPX_{i,j} - \text{median}(ExtNPX_{i, \text{Plate Controls}})\]
    • Note

      More generally, when the Plate Controls in a dataset differ from the reference Plate Control lot used by the analysis pipeline, an internal Plate Control Lot Factor can be applied to Plate Control extNPX values to align them to that reference.

      \[ExtNPX_{i, PC} (\text{adjusted}) = ExtNPX_{i, PC} (\text{raw}) + \text{PC Lot Factor}_i\] Therefore, the actual PC normalization formula is: \[NPX_{i,j} = ExtNPX_{i,j} - \text{median}(ExtNPX_{i, \text{Plate Controls}}) - \text{PC Lot Factor}_i\]

  • Intensity Normalization: Intensity normalization is a “global” adjustment that uses the actual biological samples to align plates. It is designed to further reduce technical noise and increase statistical power. \[NPX_{i,j} = ExtNPX_{i,j} - \text{median}(ExtNPX_{i, \text{Samples}})\]
    • Normalize by median of samples (excluding control strip).

---

3. Control Systems & Quality Control (QC)

The platform relies on a sophisticated hierarchy of controls to ensure data quality ExploreHT_QC.pdf.

Internal and external controls

Figure 1: Internal and external controls

The QC workflow

Figure 2: Olink QC workflow

Plate QC

• Sample QC: Samples and external controls that fail Sample QC will not be considered for additional QC steps and not normalized. Only counts will be reported for those.
• Assay QC: Detection of high number of counts for any assay, relative to the internal controls, in any of the Negative Controls is considered as unexpected signal. This step is performed on Negative Controls that pass Sample QC.

---

4. Platform Reliability: The “CV Gap”

There is a documented discrepancy between manufacturer-reported reliability and independent study results.

Olink Internal Validation

Official metrics report high precision ExploreHT_Validation.pdf: * IntraCV (Within-plate): Median ~11.2%. * InterCV (Between-plates): Median ~8.7%.

• Distribution of intra-and inter CVs

Figure 3: Distribution of intra-and inter CVs

• CIMAC_Validation.pdf 124 plasma samples (from 115 cancer patients, 5 individual health donor, 1 pooled plasma from healthy donors) were assayed using Olink Explore HT. The CV for each block was assessed by NPX values of selected (5%) of the analytes among replicates of sample controls, per Olink protocol.

Table 1: CV (%) of sample controls

Block	# of assays	Dilution factor	Intra-assay %CV mean	Inter-assay %CV mean
1	742	1:1	23.3	20.7
2	1314	1:1	13.3	11.8
3	1204	1:1	9.8	7.1
4	1106	1:1	7.2	3.5
5	582	1:10	6.6	3.8
6	270	1:100	5.6	5.3
7	134	1:1000	11.0	6.2
8	68	1:100,000	8.6	12.4

Note

A subset of 291 selected assays (~5%) that are used to assess CVs in Olink validation. This subset is based on proteins that are typically well-expressed in healthy plasma to enable the calculation of reliable CV values.

Third-Party Findings (Rooney et al., 2025)

• Independent evaluation using the ARIC cohort (102 split samples) reported lower precision Rooney2025_ARIC.pdf:
• Median CV: 35.7% for the whole Explore HT panel and 17.6% after excluding values < LOD.
• Conclusion: High variation is often driven by the large number of assays residing near the technical noise floor in clinical samples.

---

5. Handling the Limit of Detection (LOD)

The LOD is the threshold where the protein signal is statistically distinguishable from the Negative Control background.

LOD and Data Quality

Reliability is strongly tied to the signal-to-noise ratio Rooney2025_ARIC.pdf: * Precision is inversely correlated with the percentage of samples above LOD (\(r = -0.77\)). * Assays where \(NPX < LOD\) are dominated by technical noise, leading to artificially inflated CVs.

Best Practices

1. Filtering for Validation: When calculating IntraCV or InterCV, exclude data points where \(NPX < LOD\).
2. Imputation for Analysis: For biological discovery, Olink recommends original NPX. However, some researchers replace values below LOD with \(LOD/2\) to stabilize correlation analysis.
3. Reporting: Always report the “Percent Above LOD” for every assay as a primary quality metric.

---

6. Output of Olink HT

Parquet file (comparessed data frame)

Table 2: Olink NPX Explore HT Data Column Specifications

Column	Description	Type	Typical value
SampleID	The annotated sample ID	String
Sample Type	Type of sample	String	PLATE_CONTROL, NEGATIVE_CONTROL, CONTROL, SAMPLE
WellID	Id for well	String	Capital letter A–H followed by number 1–12
PlateID	Name of the plate the sample was run on	String
DataAnalysisRefID	Reference ID for data analysis	String
OlinkID	OlinkID for assay	String
UniProt	UniProt ID for assay	String
Assay	Gene name for assay	String
AssayType	Type of assay	String	Amp_ctrl, inc_ctrl, ext_ctrl
Panel	Panel name	String	Explore_HT
Block	Name of the block the sample was run on	String	1, 2, 3, 4, 5, 6, 7, or 8
Count	The total number of counts	Integer	Greater than or equal to 1
ExtNPX	Intermediate value between count and NPX: log2 of the ratio between data-point Count value and the count for the Extension Control assay for the same sample.	Double	-1.94701
NPX	NPX value	Double
Normalization	Type of normalization used in project	String	Plate control, Intensity or EXCLUDED
PCNormalizedNPX	NPX value displayed if plate control normalization has been chosen.	Double	1.735509
AssayQC	Overall QC status for an assay	String	NA, PASS, WARN
SampleQC	Overall QC status for a sample in a block	String	NA, PASS, WARN, FAIL
ExploreVersion	Software version of the module in NPX Explore HT & 3072	String

TipChoosing Between PC and Intensity Normalization

Olink provides normalized Parquet outputs upon request or based on experimental design (e.g., sample randomization). Intensity normalization is generally recommended as the primary method. This is because the standard NPX column will contain Intensity-normalized values, while PC-normalized values remain accessible via the dedicated PCNormalizedNPX column.

Olink Analysis report

• QC summary: Samples passed QC
• List of assays failed QC

Olink data analysis with R

1. Project Setup and Environment Management (with renv)

renv::restore()

- The library is already synchronized with the lockfile.

2. Load R packages and self-defined functions

• OlinkAnalyze 4.4.0 (released Dec 2, 2025) was tested in this study.

library(tidyverse)
library(magrittr)
library(OlinkAnalyze)
library(knitr)
library(kableExtra)

source("R/olink_helpers.R")

3. Load Olink data

• The two Olink parquet files are available at T-drive, and are accessible under /Volumes/DCEG/CGF/ after mounting to MacOSX (smb://gigantor.nci.nih.gov and select the volumne DCEG).
• Two Olink projects
  • RP0084-045 (rp2)
  • CS036024; Q-13387-3 (d1)

rp2_olink_fn <- "/Volumes/DCEG/CGF/Laboratory/Projects/MR-0084/RP0084-045/Data/NPXMap Exports/RP0084-045_Extended_NPX_2025-10-27.parquet"
d1_olink_fn <- "/Volumes/DCEG/CGF/Laboratory/Projects/DESL Aliquoting Projects/NAS_CS036024/Olink_DataDelivery/Q-13387_Hutchinson_Extended_NPX_2024-06-26.parquet"

rp2_npx <- read_NPX(rp2_olink_fn)

ℹ This parquet file is for research use only:
"For Research Use Only. Not for use in diagnostic procedures."!

Multiple quantification columns detected (NPX, PCNormalizedNPX, Count). NPX will be used for downstream analysis.

ℹ Outdated Data Analysis Reference ID and Panel Archive Version combination detected.

→ Re-export data using Panel Archive Version 1.5.0+ anduse the newest version of the Fixed LOD file when calculating LOD (Version 6+).

! Failure to re-export may result in incorrect PC normalization across lots and Fixed LOD calculations.

dim(rp2_npx)

[1] 783360     30

rp2_npx %>%
  head() %>%
  kable() %>%
  kable_styling(bootstrap_options = c("striped", "hover"), 
                full_width = FALSE) %>%
  scroll_box(width = "100%")

SampleID	SampleType	WellID	PlateID	DataAnalysisRefID	OlinkID	UniProt	Assay	AssayType	Panel	Block	Count	ExtNPX	NPX	Normalization	PCNormalizedNPX	AssayQC	SampleQC	SoftwareVersion	SoftwareName	PanelDataArchiveVersion	PreProcessingVersion	PreProcessingSoftware	InstrumentType	IntraCV	InterCV	SampleBlockQCWarn	SampleBlockQCFail	BlockQCFail	AssayQCWarn
IO_3862_1158_1	SAMPLE	A7	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	17461	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0
IR_8487_1158_1	SAMPLE	A8	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	16613	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0
KF_0956_1158_1	SAMPLE	A9	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	2220	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0
UM_0172_1035_1	SAMPLE	A10	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	14407	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0
IA_7060_1158_1	SAMPLE	A11	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	1231	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0
Sample_12	SAMPLE_CONTROL	A12	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	17385	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	0	1	1	0

# drive sample information from the parquet data frame
rp2_sam <- rp2_npx %>% select(1:4) %>%
  unique %>% 
  rename(plate = PlateID, well = WellID) %>%
  mutate(column = paste0("Column ", substr(well, 2, 3))) %>%
  mutate(row = substr(well, 1, 1))

rp2_sam 

Tip783,360 rows in the data frame rp2

There are 144 samples in the project rp2 and 5440 assays for each sample.

5440*144 = 783360

4. Plot Olink plates

OlinkAnalyze::olink_displayPlateLayout(data = rp2_sam, fill.color="SampleType")

Figure 4: Olink late plots of rp2

5. Olink QC

• In SampleQC column:
  • “PASS”: Datapoints from samples that pass all Sample QC criteria.
  • “FAIL”: Datapoints from samples that fail in any of the Sample QC criteria.
  • “WARN”: Datapoints from samples that get warning in any of the QC criteria.
  • “NA”: Datapoints related to an excluded assay.
• In AssayQC column:
  • “PASS”: Datapoints from assays that pass the assay QC criteria.
  • “WARN”: Datapoints from assays that get warning in assay QC.
  • “NA”: Datapoints related to excluded assays, internal controls, or assays where QC cannot be performed due to failed negative controls.

Function Definition: inspect_olink_qc

The OlinkAnalyze package contains an internal function, npxCheck(), which validates NPX data integrity and identifies cases where entire samples or assays consist of missing values (NA). However, it does not provide granular details regarding Sample-Block QC status, nor does it explicitly report on assays or samples flagged with a “WARN” status. To address this, we define the custom function inspect_olink_qc. This function extends standard validation by summarizing Olink QC flags across specific blocks, providing a clearer overview of data quality for downstream analysis.

inspect_olink_qc <- function(dat) {
  
  # 1. Identify Assays that are not "PASS" (excluding controls)
  # Filtering for AssayType=="assay" ensures we focus on actual targets
  flagged_assays <- dat |> 
    filter(AssayType == "assay") |> 
    select(OlinkID, Assay, Block, AssayQC) |> 
    distinct() |> 
    filter(AssayQC != "PASS")
  
  # 2. Identify unique SampleIDs that are not "PASS" or "NA"
  # Based on Olink definitions, "NA" usually refers to excluded assays, not failed samples
  flagged_sample_ids <- dat |> 
    filter(!SampleQC %in% c("NA", "PASS")) |> 
    pull(SampleID) |> 
    unique()
  
  # 3. Create a block-wise matrix of SampleQC status for flagged samples
  # This helps visualize if a sample failed across all blocks or just one
  flagged_samples_matrix <- dat |> 
    filter(SampleID %in% flagged_sample_ids) |>
    group_by(SampleID, Block, SampleQC) |> 
    summarise(N = n(), .groups = 'drop') |>
    mutate(status_label = paste0(SampleQC, " (n=", N, ")")) |>
    group_by(SampleID, Block) |> 
    summarise(QC = stringr::str_flatten(status_label, collapse = ", "), .groups = 'drop') |> 
    tidyr::pivot_wider(names_from = Block, values_from = QC)
  
  # Return results as a named list
  list(
    flagged_assays = flagged_assays, 
    flagged_sample_ids = flagged_sample_ids,
    flagged_samples_matrix = flagged_samples_matrix
  )
}

TipInternal controls have AssayQC as “NA”

filter(AssayType == “assay”) is applied to exclude all internal controls.

qc_results <- inspect_olink_qc(rp2_npx)

qc_results$flagged_assays
qc_results$flagged_samples_matrix

Table 3: Comprehensive QC Results

(a) Assay-level Flags

(b) Sample-level Matrix

QC Summary

ImportantQC Summary: Flagged Assays and Samples

• Quality control analysis of the rp2_npx dataset identified 16 flagged assays and six problematic samples across eight blocks. The assay-level QC revealed that 15 proteins (spread across Blocks 2, 3, 4, and 5) were categorized as “NA,” indicating they were excluded from standard processing, while one assay (C6orf118 in Block 2) received a “WARN” status.
• Sample-level analysis showed two distinct groups of performance: the “IA” group (IA_7060_1158_1/2) consistently passed most blocks but triggered “WARN” flags in Block 3, whereas the “IR” group (IR_4935_1158_1/2/3/4) exhibited severe failures, specifically a complete “FAIL” in Block 1 and varying degrees of “FAIL” or “WARN” across Blocks 2, 3, and 4. While all samples performed well in Blocks 5, 6, 7, and 8, the widespread issues in the first four blocks—particularly for the “IR” samples—suggest these data points should be handled with caution or excluded to maintain study integrity.

---

6. Visualizations

Heatmap

### fontsize_row does not work so I have to change fontsize 
olink_heatmap_plot(rp2_npx, variable_row_list =  'SampleType', variable_col_list = "Block", show_colnames=F, cluster_cols=F, center_scale=T, fontsize=6, fontsize_row = 2)

Warning: The assays OID40764, OID40857, OID42179, OID42282, OID42294, OID42327,
OID42394, OID42606, OID43241, OID43371, OID43738, OID43750, OID44351, OID44410,
OID44923 have NPX = NA for all samples. They will be excluded from the analysis

1 assay(s) exhibited assay QC warning. For more information see the AssayQC column.

Figure 5: Olink NPX Heatmap

TipNotes on Heatmap Generation

• Data Scaling: By default, NPX values are centered and scaled (Z-score standardized) across assays.
• Assay Filtering: 15 assays were excluded due to an AssayQC status of FAIL.
• Sample Quality: NPX values for samples with a SampleQC status of FAIL are set to NA. These are represented by black bars in the heatmap.

Generates boxplots of NPX vs. SampleID using olink_dist_plot

• We assigned pandel value by PlateID here

Boxplots of NPX is another way to visualization the distribution of NPX values in each sample.

p <- rp2_npx |>
  olink_dist_plot(color_g = "SampleQC") + 
  facet_grid(Block ~ PlateID, scales = "free_x", space = "free_x") + 
  theme(axis.text.x = element_text(angle = 90, size = 8, vjust = 0.5, hjust = 1))

Warning: The assays OID40764, OID40857, OID42179, OID42282, OID42294, OID42327,
OID42394, OID42606, OID43241, OID43371, OID43738, OID43750, OID44351, OID44410,
OID44923 have NPX = NA for all samples. They will be excluded from the analysis

1 assay(s) exhibited assay QC warning. For more information see the AssayQC column.

p$layers[[1]]$geom_params$outlier_gp$alpha <- 0.5
p$layers[[1]]$geom_params$outlier_gp$size <- 0.5
p$layers[[1]]$geom_params$outlier_gp$shape <- 21
p$layers[[1]]$geom_params$outlier_gp$colour <- "transparent"
p$layers[[1]]$geom_params$outlier_gp$fill <- "black"

p

Figure 6: NPX distribution by panel

olink_qc_plot

Generates a facet plot per Panel using ggplot2::ggplot and ggplot2::geom_point and stats::IQR plotting IQR vs. median for all samples. Horizontal dashed lines indicate +/-IQR_outlierDef standard deviations from the mean IQR (default 3). Vertical dashed lines indicate +/-median_outlierDef standard deviations from the mean sample median (default 3).

### Negative controls should be excluded
qc <- olink_qc_plot(rp2_npx |> filter(SampleType=="SAMPLE"), color_g = "PlateID", IQR_outlierDef = 3, median_outlierDef = 3)

Warning: The assays OID40764, OID40857, OID42179, OID42282, OID42294, OID42327,
OID42394, OID42606, OID43241, OID43371, OID43738, OID43750, OID44351, OID44410,
OID44923 have NPX = NA for all samples. They will be excluded from the analysis

1 assay(s) exhibited assay QC warning. For more information see the AssayQC column.

qc 

outliers <- qc$data %>% filter(Outlier == 1)
outliers

Figure 7: Olink QC plot

Figure 8: Olink QC plot

PCA plot

olink_pca_plot(rp2_npx |> filter(SampleType=="SAMPLE"), color_g = "SampleQC")

Warning: The assays OID40764, OID40857, OID42179, OID42282, OID42294, OID42327,
OID42394, OID42606, OID43241, OID43371, OID43738, OID43750, OID44351, OID44410,
OID44923 have NPX = NA for all samples. They will be excluded from the analysis

1 assay(s) exhibited assay QC warning. For more information see the AssayQC column.

Warning in npxProcessing_forDimRed(df = df, color_g = color_g, drop_assays =
drop_assays, : There are 4356 assay(s) that were imputed by their medians.

olink_pca_plot(rp2_npx |> filter(SampleType=="SAMPLE"), color_g = "PlateID")

Warning: The assays OID40764, OID40857, OID42179, OID42282, OID42294, OID42327,
OID42394, OID42606, OID43241, OID43371, OID43738, OID43750, OID44351, OID44410,
OID44923 have NPX = NA for all samples. They will be excluded from the analysis

1 assay(s) exhibited assay QC warning. For more information see the AssayQC column.

Warning in npxProcessing_forDimRed(df = df, color_g = color_g, drop_assays =
drop_assays, : There are 4356 assay(s) that were imputed by their medians.

(a) Color by SampleQC

(b) Color by PlateID

Figure 9: Olink PCA plot

UMAP plot

olink_umap_plot(rp2_npx |> filter(SampleType=="SAMPLE"), color_g = "SampleQC")

Loading required namespace: umap

Warning: The assays OID40764, OID40857, OID42179, OID42282, OID42294, OID42327,
OID42394, OID42606, OID43241, OID43371, OID43738, OID43750, OID44351, OID44410,
OID44923 have NPX = NA for all samples. They will be excluded from the analysis

1 assay(s) exhibited assay QC warning. For more information see the AssayQC column.

Warning in npxProcessing_forDimRed(df = df, color_g = color_g, drop_assays =
drop_assays, : There are 4356 assay(s) that were imputed by their medians.

olink_umap_plot(rp2_npx |> filter(SampleType=="SAMPLE"), color_g = "PlateID")

Warning: The assays OID40764, OID40857, OID42179, OID42282, OID42294, OID42327,
OID42394, OID42606, OID43241, OID43371, OID43738, OID43750, OID44351, OID44410,
OID44923 have NPX = NA for all samples. They will be excluded from the analysis

1 assay(s) exhibited assay QC warning. For more information see the AssayQC column.

Warning in npxProcessing_forDimRed(df = df, color_g = color_g, drop_assays =
drop_assays, : There are 4356 assay(s) that were imputed by their medians.

(a) Color by SampleQC

(b) Color by PlateID

Figure 10: Olink UMAP plot

---

7. Integrating LOD

NPX values below Limit of Detection (LOD) are not reliable. In the OlinkAnalyze R package, the olink_lod() function provides two primary methods for integrating Limit of Detection (LOD) information into Olink Explore datasets:

• Negative Control LOD (NCLOD)

  This method calculates LOD values based specifically on the negative controls included within your own study’s dataset.

  • How it works: It uses the negative controls to determine the background noise. For assays with higher counts, it typically calculates the median NPX plus 3 standard deviations (or 0.2 NPX, whichever is larger). For assays with very low counts, it calculates LOD from the count values and converts them to NPX.
  • Requirement: Your dataset must contain at least 10 negative controls that have passed quality control (SampleQC).
  • Best For: Larger projects where you want LOD values tailored to the specific technical conditions and background of your unique experiment.

• Fixed LOD (FixedLOD)

  This method uses predetermined LOD values provided by Olink, rather than calculating them from your current data.

  • How it works: You must download a “Fixed LOD” reference file (CSV) from the Olink Document Download Center. These values are generated by Olink when a new reagent lot is released.
  • Requirement: You must provide the path to this external CSV file in the lod_file_path argument of the function.
  • Best For: Smaller studies (fewer than 10 negative controls) where there isn’t enough internal data to calculate a statistically reliable project-specific LOD.

The latest fixed LOD file is available for download here (see Table 4 (a)). This file contains reference data for DataAnalysisRefID (DarID) DX00YY, where X represents blocks 1–8 and YY indicates the reference version (currently ranging from 01 to 14). In the rp2 study, the DataAnalysisRefID (or data_analysis_ref_id) is designated as D10007 through D80007.

The detailed use of olink_lod() is described here.

Append fixed LOD

As there are no sufficient negative controls included in rp2, we can only apply fixed LOD here.

fixedLOD_fn <- "data/Explore HT_Fixed LOD.csv"
fixedlod.dat <- read.delim2(fixedLOD_fn, sep=';')

# Use kable to ensure Quarto treats these as separate citable tables
knitr::kable(head(fixedlod.dat))
rp2_lod <- rp2_npx |> 
  olink_lod(lod_file_path = fixedLOD_fn, lod_method = "FixedLOD")

Warning: The assays OID40764, OID40857, OID42179, OID42282, OID42294, OID42327,
OID42394, OID42606, OID43241, OID43371, OID43738, OID43750, OID44351, OID44410,
OID44923 have NPX = NA for all samples. They will be excluded from the analysis

1 assay(s) exhibited assay QC warning. For more information see the AssayQC column.

knitr::kable(head(rp2_lod))

Table 4: Olink Explore HT Fixed LOD Analysis

(a) Fixed LOD downloaded from Olink website

OlinkID	AssayType	UniProt	Assay	Panel	Block	DataAnalysisRefID	BimodalDistribution	LODNPX	LODCount	LODMethod	Version
OID40001	assay	Q9NUQ8	ABCF3	Explore_HT	1	D10001	FALSE	0.976509184936807	2086	lod_npx	9.0.0
OID40002	assay	Q8WTS1	ABHD5	Explore_HT	1	D10001	FALSE	2.20770109787743	564	lod_npx	9.0.0
OID40003	assay	Q96I13	ABHD8	Explore_HT	1	D10001	FALSE	6.22270040477226	150	lod_count	9.0.0
OID40004	assay	Q9H845	ACAD9	Explore_HT	1	D10001	FALSE	7.2585583689779	178	lod_count	9.0.0
OID40005	assay	Q6NUN0	ACSM5	Explore_HT	1	D10001	FALSE	2.67065654184691	572	lod_npx	9.0.0
OID40006	assay	Q8IZF7	ADGRF2	Explore_HT	1	D10001	FALSE	4.91106891907256	386	lod_npx	9.0.0

(b) Olink NPX data with integrated LOD (LOD and PCNormalizedLOD columns appended)

SampleID	SampleType	WellID	PlateID	DataAnalysisRefID	OlinkID	UniProt	Assay	AssayType	Panel	Block	Count	ExtNPX	NPX	Normalization	PCNormalizedNPX	AssayQC	SampleQC	SoftwareVersion	SoftwareName	PanelDataArchiveVersion	PreProcessingVersion	PreProcessingSoftware	InstrumentType	IntraCV	InterCV	SampleBlockQCWarn	SampleBlockQCFail	BlockQCFail	AssayQCWarn	LOD	PCNormalizedLOD
IO_3862_1158_1	SAMPLE	A7	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	17461	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0	NA	NA
IR_8487_1158_1	SAMPLE	A8	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	16613	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0	NA	NA
KF_0956_1158_1	SAMPLE	A9	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	2220	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0	NA	NA
UM_0172_1035_1	SAMPLE	A10	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	14407	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0	NA	NA
IA_7060_1158_1	SAMPLE	A11	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	1231	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	1	1	1	0	NA	NA
Sample_12	SAMPLE_CONTROL	A12	RP0084-045_A	D10007	OID45511	EXT1	Extension control 1	ext_ctrl	Explore_HT	1	17385	0	0	Intensity	0	NA	PASS	1.3.0	NPX Map	1.3.0	5.1.0	ngs2counts	Illumina NovaSeq X Plus	NaN	NaN	0	1	1	0	NA	NA

Analysis of RP2 Dataset: Assays with NPX Values Below LOD

sample_na_cnts <- rp2_lod |>
  filter(AssayType=="assay") |>
  group_by(SampleID, SampleType, PlateID) |>
  summarise(
    N=sum(NPX < LOD, na.rm=T),
    total_assays=n(),
    non_nas = sum(!is.na(NPX)),
    .groups="drop"
  ) |> 
  group_by(PlateID) |>
  summarise(
    Total=first(total_assays),
    NonNA=first(non_nas),
    Q1=quantile(N, 0.25, na.rm=T),
    Median_Below_LOD=quantile(N, 0.50, na.rm=T),
    Q3=quantile(N, 0.75, na.rm=T),
    .groups="drop"
  )
  
sample_na_cnts

Table 5: Assays with NPX Values Below LOD

---

8. Advanced analysis

Load more sample information

library(readxl)
.subj <- read_excel("data/Olink HCC Sample Details.xlsx", skip=1, sheet=1) |>
  select(1:6) |>
  setNames(c("subj_id", "aliquot_id", "subsample_id", "subj_type", "sex", "tissue_type"))

New names:
• `Subject` -> `Subject...1`
• `` -> `...13`
• `Subject` -> `Subject...14`

.sam <- read_excel("data/Olink HCC Sample Details.xlsx", sheet=4) |>
  setNames(c("project", "plate_id", "row", "col", "well", "subj_id", "aliquot_id", "subsample_id", "olink_sample_id")) |>
  select(-row, -col, -well)

### We expect there are 48 samples in CS036024 (Q-13387-3) and 124 samples in RP0084-045
.sam %$% table(project)

project
CS036024 (Q-13387-3)           RP0084-045 
                  48                  124 

### find out how many CGR samples matched in d1 and rp2

rp2_sam_full <- rp2_sam |>
  inner_join(.sam, by=join_by(SampleID == olink_sample_id)) |>
  inner_join(.subj |> select(subj_id, subj_type, sex, tissue_type))

Joining with `by = join_by(subj_id)`

### 20 external contorl samples are not included.  So there are only 124 rows
rp2_sam_full

Table 6: Detailed sample information of rp2

Customized PCA plot

my_olink_pca(all_fix %>% filter(SampleID %in% cgr_all$SampleID), cgr_all, col_by=“subj_type”, panel_by=“plate”, ncol=1) ggsave(“graphs/Q7.my_pca.all_fix.type.pdf”, width=10, height=8)

run_olink_pca_workflow(rp2_npx %>% filter(SampleID %in% rp2_sam_full$SampleID), rp2_sam_full, color_by="plate", facet_by = NULL) 

run_olink_pca_workflow(rp2_npx %>% filter(SampleID %in% rp2_sam_full$SampleID), rp2_sam_full, color_by="subj_type", facet_by="plate") 

run_olink_pca_workflow(rp2_npx %>% filter(SampleID %in% rp2_sam_full$SampleID), rp2_sam_full, color_by="sex", facet_by="plate") 

(a) Baseline version (reproduced from Figure 9 (b))

(b) Stratified by Subject Type

(c) Stratified by Sex

Figure 11: Customized PCA plot analysis

Intra-CV analysis

• Based on the Olink Explore HT sample controls, 291 selected assays yielded computable intra- and inter-assay CV values.
• We defined a function calc_intraCV_nse to calculate IntraCV for any replicated samples including those Olink SAMPLE_CONTROL.
  • InterCV can be calcuated in a similar way.

1. Compute Intra-CV for rp2

rp2_npx |> 
  group_by(PlateID, SampleID) |>
  summarise(
    N_IntraCV=sum(!is.na(IntraCV)),
    N_InterCV=sum(!is.na(InterCV))
  )

`summarise()` has grouped output by 'PlateID'. You can override using the
`.groups` argument.

### To better evaludate Olink HT performance, we managed to calculate intra- and inter-CV of all replicated samples, including Sample Controls
rp2_intraCV <- calc_intraCV_nse(rp2_lod, rp2_sam_full, rp2_sam)

### Our calculation should be matched with the Olink intraCV in those 291 assays
.dat <- rp2_intraCV |>
  filter(PlateID=="PlateLayout_RP0084-045_B_ExploreHT (3)" & subsample_id == "SAMPLE_CONTROL") |>
  select(OlinkID, newIntraCV=IntraCV) |>
  inner_join(
    rp2_npx |> 
      filter(PlateID=="PlateLayout_RP0084-045_B_ExploreHT (3)" & !is.na(IntraCV))|>
      select(OlinkID, IntraCV) |> 
      unique() # all samples have the same intraCV in the plate
  )

Joining with `by = join_by(OlinkID)`

### all matched
.dat %$% table(newIntraCV - IntraCV < 1e-5, useNA="ifany")

TRUE 
 291 

2. Get the median of IntraCV of each assay

rp2_intraCV.median <- rp2_intraCV |> 
  mutate(
    SampleType = ifelse(grepl("_CONTROL",subsample_id), subsample_id, "SAMPLE")
  ) |> 
  filter(AssayType == "assay" ) |>
  group_by(OlinkID, PlateID, SampleType) |> 
  summarise(IntraCV_median = median(IntraCV, na.rm=T), .groups="drop")

3. Summary median Intra-CV by SampleType for each plate

rp2_intraCV.sum_all <- rp2_intraCV.median |>
  group_by(PlateID, SampleType) |> 
  summarise(
    N=n(),
    Q1=quantile(IntraCV_median, 0.25, na.rm=T),
    Q2=quantile(IntraCV_median, 0.50, na.rm=T),
    Q3=quantile(IntraCV_median, 0.75, na.rm=T),
    .groups="drop"
  )


# Use kable to render the data frame as a formal table
knitr::kable(rp2_intraCV.sum_all)

Table 7: IntraCV with all NPX

PlateID	SampleType	N	Q1	Q2	Q3
PlateLayout_RP0084-045_B_ExploreHT (3)	NEGATIVE_CONTROL	5401	9.691049	23.57762	52.62142
PlateLayout_RP0084-045_B_ExploreHT (3)	PLATE_CONTROL	5401	8.473897	16.10635	31.50751
PlateLayout_RP0084-045_B_ExploreHT (3)	SAMPLE	5401	9.623662	17.26379	28.80782
PlateLayout_RP0084-045_B_ExploreHT (3)	SAMPLE_CONTROL	5401	7.073281	13.69899	26.84559
RP0084-045_A	NEGATIVE_CONTROL	5401	10.222677	23.51408	51.76379
RP0084-045_A	PLATE_CONTROL	5401	7.913498	15.44079	29.94569
RP0084-045_A	SAMPLE	5401	8.084016	14.22349	25.89999
RP0084-045_A	SAMPLE_CONTROL	5401	7.609023	14.69867	29.18682

4. Make plots as Rooney et al, 2025

### Fig1
rp2_intraCV.median |>

  mutate(IntraCV_median=ifelse(IntraCV_median> 50, 50, IntraCV_median)) |>
  ggplot(aes(x=IntraCV_median, fill=PlateID)) +
    geom_density( alpha=0.2) +
    facet_grid(. ~ SampleType) + 
    theme(legend.position = "bottom")

### Fig 2
.dat <- rp2_lod |> 
  filter(SampleType=="SAMPLE") |>
  group_by(OlinkID, PlateID) |>
  summarise(
    Perc_gtLOD = mean(NPX>LOD, na.rm=T),
    .groups='drop'
  ) |>
  inner_join(
      rp2_intraCV.median |> filter(SampleType == "SAMPLE")
  )

Joining with `by = join_by(OlinkID, PlateID)`

# Assuming your data is in a data frame named 'df'
ggplot(data = .dat, aes(x = Perc_gtLOD, y = IntraCV_median, col=PlateID)) +
  geom_point(size = 1, alpha = 0.8) +
  geom_smooth(method = "lm", 
              formula = y ~ x, 
              color = "black", 
              linetype = "dashed", 
              linewidth = 0.8,
              # This line is crucial: the plot description mentions the CV 
              # was log2 transformed for the regression.
              # However, since we are showing the line on a log10 axis, 
              # it is best to let ggplot handle the scale transformation 
              # or ensure the CV data is already log2 transformed 
              # for the fit, which is complex. 
              # A simpler interpretation is to fit the line to the data shown on the plot:
              se = FALSE) +
  scale_y_log10() + facet_wrap(~PlateID) + theme(legend.position = "bottom")

(a) Distribution of IntraCV (similar to Fig 1A. in [Rooney et al, 2025])

(b) Scatterplot of the percentage of participants with protein values above the LOD vs CV (displayed on the log2 scale) on the Olink Explore HT (similar to Fig 2. in [Rooney et al, 2025])

Figure 12: IntraCV with all NPX

5. Filtering NPX < LOD to improve intra-CV

As demonstrated in Figure 12 (b), the precision (CV) of Olink Explore HT assays is strongly inversely correlated with protein detectability (the percentage of samples above the validation LOD). To improve data quality, we recoded NPX values below the LOD as NA. We then generated a summary table of the intra-CV, mirroring the previous analysis where all NPX values were included, to evaluate the resulting improvement in precision.

rp2_intraCV_lod <- calc_intraCV_nse(rp2_lod, rp2_sam_full, rp2_sam, treat_LOD_as_NA = T)

rp2_intraCV.median_lod <- rp2_intraCV_lod |> 
  mutate(
    SampleType = ifelse(grepl("_CONTROL",subsample_id), subsample_id, "SAMPLE")
  ) |> 
  filter(AssayType == "assay" ) |>
  group_by(OlinkID, PlateID, SampleType) |> 
  summarise(IntraCV_median = median(IntraCV, na.rm=T), .groups="drop")
  
  
rp2_intraCV.sum_lod <- rp2_intraCV.median_lod |>
  group_by(PlateID, SampleType) |> 
  summarise(
    N=n(),
    Q1=quantile(IntraCV_median, 0.25, na.rm=T),
    Q2=quantile(IntraCV_median, 0.50, na.rm=T),
    Q3=quantile(IntraCV_median, 0.75, na.rm=T),
    .groups="drop"
  )


# Use kable to render the data frame as a formal table
knitr::kable(rp2_intraCV.sum_lod)

Table 8: IntraCV with NPX \(\ge\) LOD

PlateID	SampleType	N	Q1	Q2	Q3
PlateLayout_RP0084-045_B_ExploreHT (3)	NEGATIVE_CONTROL	750	6.580434	14.896913	30.78217
PlateLayout_RP0084-045_B_ExploreHT (3)	PLATE_CONTROL	2672	5.659866	8.527416	13.27701
PlateLayout_RP0084-045_B_ExploreHT (3)	SAMPLE	4017	7.297420	9.780197	14.18543
PlateLayout_RP0084-045_B_ExploreHT (3)	SAMPLE_CONTROL	2815	5.034002	8.330916	14.09463
RP0084-045_A	NEGATIVE_CONTROL	670	6.816213	14.583087	29.06493
RP0084-045_A	PLATE_CONTROL	2657	5.143483	8.083143	13.17757
RP0084-045_A	SAMPLE	3482	5.731077	8.538082	13.35158
RP0084-045_A	SAMPLE_CONTROL	2799	5.348042	8.918910	15.37332

Note

Note that assay precision (Intra-CV) is markedly higher after filtering for detectability (Table 8) compared to the full dataset (Table 7)